Energy & Infrastructure Insights: Will Data Center and Energy Capacity Growth Meet the Demands of AI?

As AI technologies develop apace, the question looms whether data centers and the electricity needed to power them will be able to meet the rocketing demand.

During the latest Latham Infra Circle event, “The Evolution of AI Infrastructure in Germany,” Latham partner Alexander “Stefan” Rieger spoke with David Howson (president, EMEA at Vantage Data Centers) and Aroosh Thillainathan (group founder, CEO, and Chairman at Northern Data Group) about the opportunities for data center operators and investors to meet the growing demand in Germany.

Project Stargate is a $500 billion investment initiative focused on AI infrastructure within the United States, particularly targeting data centers. Is there a need for a similar initiative, Project Stargate 2.0, in Germany?

Thillainathan: 100% yes. We must ensure that we do not fall too far behind the US and China, which are currently leading the field. I am convinced that this infrastructure is of critical importance and will only grow in significance in the future. On a positive note, Germany has already made some early-stage investments in this area. However, considering this is a $500 billion investment, we would expect a project like this to require approximately 50 gigawatts of energy per data center to reach fruition. My primary concern is around this large energy demand, but I am confident that we have enough talented individuals here to make this happen. It highlights the imperative to establish our own energy clusters in Europe.

From the perspective of operators, what trends have you observed in AI infrastructure demand over the past few years, and what are the primary drivers behind this demand?

Howson: The demand has been quite volatile. The AI ecosystem is still in its nascent stages, yet it is evolving at an incredibly rapid pace — likely three to five times faster than the early cloud adoption cycle from 10 to 15 years ago. Those involved in data center development are aware that development cycles are becoming longer, primarily due to challenges in securing power access, as well as obtaining environmental and planning permits. There are numerous external factors and headwinds to contend with. These businesses are essentially making predictions about the capacity they will require five to seven years down the line, which is a daunting task. While there haven’t been many significant inflection points in terms of the requirements for power, cooling, and security, we did experience such a shift last year.

"The AI ecosystem is still in its nascent stages, yet it is evolving at an incredibly rapid pace — likely three to five times faster than the early cloud adoption cycle from 10 to 15 years ago."

David Howson, president (EMEA) at Vantage Data Centers

What role do hyperscalers play in the AI infrastructure wave, and how do you anticipate their demands will evolve over time?

Howson: Currently, hyperscalers are the primary drivers of demand. However, I anticipate a significant influx of new entrants — including new hyperscalers — in the near future, whose development will be interesting to watch.

Do you believe the German power market is adequately prepared for the AI infrastructure wave?

Thillainathan: From a German or broader European energy perspective, I have some concerns. Meeting the rising energy demand from the AI sector in Europe will be challenging, as we anticipate the need to triple capacity by 2030 to nearly 150 terawatt hours (TWh). This represents a substantial increase of our energy requirement, and I believe it will pose a significant challenge for our energy market.

"Meeting the rising energy demand from the AI sector in Europe will be challenging, as we anticipate the need to triple capacity by 2030 to nearly 150 terawatt hours (TWh)."

Aroosh Thillainathan, group founder, CEO, and Chairman at Northern Data Group

Is this increasing demand for electricity compatible with sustainability objectives?

Howson: For AI workloads to coexist with cloud applications and other services, the required density per kilowatt per square meter is expected to double, triple, or even quadruple, based on current chip technology and electrical needs. However, I am optimistic that chip technology will become more effective and efficient over time, with significant advancements on the horizon. We are all striving to find a balance to avoid over-investing in electrical infrastructure that may not be necessary five years from now. From a sustainability standpoint, there is a tremendous opportunity to return power to the grid during off-peak times or to leverage advanced technologies, including battery technologies. I anticipate the adoption of many efficient power and cooling technologies that have been somewhat on the periphery but are yet to be fully utilized in data centers.

To what extent are your data centers powered by renewable energy, and do you have a strategy to achieve carbon neutrality in the coming years?

Thillainathan: We own the majority of our data centers, and our portfolio is primarily located in Scandinavian countries, where access to renewable energy is excellent. However, internationally, it can be more challenging. In the US, for instance, the availability of renewable energy depends on the specific region, making it difficult to ascertain the exact energy mix. Nevertheless, we are committed to selecting the right locations and partners to advance our effort to maintain carbon neutrality as much as possible, and our aim as a business is to use renewable energy wherever it is available.

If an investor were to offer you €1 billion, in which country would you choose to build a data center, and why?

Thillainathan: The decision largely hinges on the time to market. In the US, we can have a project up and running within two years. Over the past four years, our attempts to establish a data center in Germany have proved to be extremely challenging. While there are positive developments, I currently don’t see an environment in Germany that could match the US.

Howson: The new US administration is actively reducing bureaucratic hurdles. There was already less red tape in the US, and any remaining obstacles are being swiftly removed to capture the market rapidly.

What are the biggest drivers of change likely to be?

Howson: A significant driver will likely stem from consumer use, particularly the concept of a personal AI that evolves with you as a kind of assistant. I believe this will be a major use case for AI. The security and data storage complexities that emerged with cloud and personal data will be elevated to an entirely new level.

Would it make more sense to relocate data centers to where the power is, rather than where the customers are?

Howson: It’s a balance between economics and technology. Certain applications have latency and performance constraints that necessitate proximity to customers. However, there may be scenarios where power is brought closer to customers. I’ve observed significant investments in new technologies, such as small modular reactors, but ultimately, it all hinges on regulatory approvals.